17

I'm frequently getting 408 Request Time-out errors when browsing through Stack Overflow and Meta. I used to get this four or five times a day before, but now it has become very frequent. I've read this February 2011 question in which Kyle Brandt ♦ says:

We send our web logs to SQL so we can query them. It turns out that we have a bug that prevents 408s from being logged to SQL. [...] I will do more extensive research when our 408s are part of queryable logs (should be a week or two [...]).

My internet connection is somewhat stable and I haven't experienced this issue on other sites before. I don't think it's a browser-specific issue as it is reproducible on latest versions of Google Chrome and Firefox (on Ubuntu 13.10). However, it is specific to Stack Overflow and Meta. I'm unable to reproduce this on other SE sites.

Am I the only one facing this issue? Could it be something wrong with the regional CDN or something like that? Could someone please take a look at this issue? (not pointing hands)

This appears when I try to close-vote questions, too:

10
  • I have seen these more often since I've moved to a place with a less-than-perfect WLAN. It's not often enough for me to be annoying, though.
    – Pekka
    Nov 24, 2013 at 18:29
  • I see this sometimes, rarely, after a couple second delay on a bad connection.
    – user206222
    Nov 24, 2013 at 18:38
  • 3
    I've seen that when I was having network issues, but never when my network connection works perfectly Nov 24, 2013 at 18:44
  • I get that error regularly with weak edge mobile connections.
    – rekire
    Nov 24, 2013 at 20:34
  • I don't think it's a bug in Stack Exchange or that there's anything the team can do, it's more likely local issue with your connection and/or ISP. Nov 26, 2013 at 12:32
  • @ShaWizDowArd: What about this? Nov 26, 2013 at 12:37
  • @AmalMurali feel free to ping Kyle by posting comment on his answer, he still work for SE and still around. (well, 6 days ago so you might have to wait couple of days) Nov 26, 2013 at 12:40
  • 2
    Upvoted: I hit this error twice this morning in quick succession, on my home cable modem connection which has been otherwise very stable this morning.
    – qwrrty
    Nov 28, 2013 at 16:17
  • 1
    See also Kyle's update: "It is possible this related to tcp-reconnect in modern browsers and our client timeout settings in haproxy. I'm going to have our latest SRE member Shane Madden investigate this now that I have a reasonable hypothesis."
    – Arjan
    May 26, 2014 at 21:08
  • Shane is asking: "Is anyone seeing these 408s consistently on a good internet connection, and not using Chrome?"
    – Arjan
    May 27, 2014 at 5:10

1 Answer 1

5
+100

As far as we can tell, the recent reports of this are due to a Chrome update.

A client side timeout seems to have stopped working, which created a small window in which a 408 sent to close a speculative connection by the server was unintentionally read as the response when attempting to use the connection to send a request.

So, you would potentially get a 408 when you sent a request between 15 and 16 seconds after your most recent request.

We've filed a bug report for Chrome, and implemented the workaround from HAProxy here.

Let me know if anyone's still observing regular 408 errors.

1
  • 2
    From your bug report: "[Chrome] will send the request after the 408 response has been sitting in the receive buffer, and use that 408 response as the apparent response to the request". Very nice find!
    – Arjan
    May 27, 2014 at 22:03

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .